jay vi

Hello everyone 👋 I've been posting (and replying!) here for months while building a unified inbox dashboard for businesses! ... And after around 6 months of development, and around 3 months of building in public - It's finally released! 🎉 It solves a problem many businesses have - scattered customer support across many different apps This is mainly felt by companies that deal with real life goods or services such as shops, logistics, hotels, outsourcing etc... although even small/medium size SaaS can benefit from it! These kinds of businesses often have customers trying to reach them via different chat apps: - Facebook Messenger - Whatsapp Business - Telegram - Email - Instagram DMs - Website Widget Converge is built to make that easy! By letting you connect all of those platform (and more coming soon), you can reply to customer support inquiries, and meet your customers where they are - With many useful features built-in - light analytics - user notes - lead tracking - AI response suggestions - Quick replies - Team collab and management it has a single all-inclusive plan, with a 2 week, no credit card free trial. It's been quite tough - mostly everything done by me with some QA help on the side. UI/UX, SEO, backend design, deployment and everything in-between! So I would be very grateful if you'd check it out, and let me know if you have any feedback (good or bad!) as I've spent an immense amount of time building it Thank you for reading this! Here's the product hunt release page👇 producthunt.com/products/conve…

general: "dont push yet leme test" claude: "nuclear launch in 3...2...1"

@ProtagXnist7 @kimmonismus i see engagement bait connection

@kimmonismus Used for robots stacking blocks. I don’t see the data center connection at all please explain

1/ Huge: Researchers just built an AI system that uses 100x less energy than current models - and it's actually more accurate. This could rewrite the rules of the entire AI industry. Lets dig into this 🧵

@kmcnam1 also true but Lets be honest here they take forever to run, that's the only reason i have only 1100 of them, otherwise i'd add much more

GLM5.1 can't stop impressing me with its design skills if ONLY it was a little smarter on long contexts. Just genuinely needs a few IQ points more. That's it. Opus is smarter, so I have to ask opus every time to figure out the issue and then hand it to GLM-5.1 and it does it beautifully!

@agile_memes Got a skill file for it?

Best depiction of this I’ve seen

@marmaduke091 AI can't do agent harness 99% failure rate (tested on my llama7b-q4 locally)

>AI can't do math >Look inside >Strongest model they tested was GPT-4o-mini Get this slop out of here. The models tested are so inferior to current SOTA models that the findings are completely irrelevant to the current AI landscape.

@Venkydotdev 30k x 30 days = 100k/y

500 users × $5/month × 12 months = $30,000/year

@ahmedogarei @melvynx 👍

@justjayvi @melvynx because its not reliable, each second query is llm upstream and runtime error because of high load of traffic on their cloud end, especially after hard anthropic ban of openclaw use

Search the Opus 4.6 replacement for my OpenClaw: day 1 I tried Kimi 2.5 today. It's not good. It has issues with tool calls. It outputs Chinese characters. It's bad... really bad. What's next?

@ahmedogarei @melvynx why not qwen3.6? It's specially tuned to do openclaw better

@melvynx been testing the qwen 3.5 plus today, not good as opus but way way better than gpt5.4, the next one on the list is aravhawk/qwen3.5-opus-4.6 via ollama cloud api

@melvynx GLM-5.1

@Ancap_Revolt @10x_er who is leader?

@justjayvi @10x_er Dude are you retarded? <leader>sf or sg will find anything in your codebase in milliseconds.

Yeah the future of engineering is cooked

@MarkBestForex @10x_er my codex-5.4-max model handles it well don't need no tool

@justjayvi @10x_er Existing tools have done this for ages.

@karmaco1a Hey this guy did something extremely fraudulent and illegal and unethical!!! ... I should also do it to become rich! 😂

Here come the imitators

@elkelk It should be what? pretty sure we're talking about different websites

Idk why but this bugs me X should be the home of high quality content and its leader exclusively quotes with “cool” and “interesting”

@HarshithLucky3 I'm sure they don't have this question, and 1000 variations of it in the training set by now there's no way that'd be true it must be the agi

Gemma 4 31B vs Qwen 3.5 27B Gemma: If you walk, you will arrive at the car wash, but your car will still be 50 meters away! Gemma 4 31B in Google AI Studio Qwen 3.5 27B on Qwen Chat

@thekitze defensive programming i like it, LGTM PR ready to merge

the type of shit u get when u let ai cook too long without reviewing jesus christ

@test_tm7873 That's awesome!! I've just trained my first 2.5M parameter model yesterday 😅 No idea what I'm doing yet! What are you using to train? IS it fast? Good luck!!!! 🚀

Good news. Claude love my V4 architecture. (rn V5 in training) very cool!

@DevBySami most people have ADHD line breaks drip feed content and it keeps their attention

I'm so tired of ChatGPT treating me like a retard who can't read. Who the fuck wants to read shit like this, genuinely

@PaulYacoubian you just got noted - good luck 😧 i hate this platform!

@justjayvi I was worried after i read your comment and then elon retweeted this one today 😂

jensen huang, founder ceo of nvidia, in 2011